IET Intelligent Transport Systems

Research Article

ISSN 1751-956X
Collecting smart meter data via public Received on 21st March 2015
Revised on 17th February 2016
transportation buses Accepted on 9th May 2016
doi: 10.1049/iet-its.2015.0058
www.ietdl.org

Bilal Erman Bilgin 1, Selcuk Baktir 1 ✉, Vehbi Cagri Gungor 2

1
Department of Computer Engineering, Bahcesehir University, Istanbul, Turkey
2
Department of Computer Engineering, Abdullah Gul University, Kayseri, Turkey
✉ E-mail: [email protected]

Abstract: With advances in technology, wireless sensor networks (WSNs) have found new applications and their
popularity has increased dramatically. In several applications, WSNs have the potential to replace wired data
communication systems, e.g. in vehicular ad hoc networks (VANETs) they are the natural option for data
communication. WSNs are also proposed for data communication in the emerging smart grid. In this study, the
authors merge these two application domains, i.e. VANET and smart grids, and propose a novel solution for effective
smart grid data communication. The authors’ proposed scheme achieves the task of collecting data from smart meters
by utilising VANETs. Using network simulator-2 and with different routing protocols, the authors have performed
simulations and shown the efficacy of their scheme in terms of average end-to-end delay and delivery ratio.

1 Introduction IEEE 802.11p allows communication in ranges up to 1000 m,

smart meters within 1000 m of a public bus stop can transmit their
Recently, with new advances, wireless sensor networks (WSNs) data to the bus stop, and buses passing by the bus stop can receive
have started replacing wired systems in several settings, including the data and then transmit it to a central database via road side units.
military, environmental monitoring, health-related and home Main Contributions:
applications, smart grid environments and vehicular
communication [1]. WSNs have been deployed widely due to their (i) There exists no published work on collecting smart meter data
low-cost and collaborative nature. by using buses or other public transportation vehicles; this is the first
The smart grid is the evolving and modernised version of the such scheme to the best of our knowledge.
existing traditional electricity grid. The main functionality of the (ii) The proposed system only needs smart meters with IEEE
smart grid is its two-way communication capability, which 802.11p communication capability and transceivers on bus stops.
facilitates self-healing through real-time monitoring and timely We propose the use of IEEE 802.11p for smart grid data collection
reaction to incidents. Furthermore, by recording consumer for the first time in literature.
electricity consumption data frequently and transmitting it to the (iii) Our proposed scheme is novel in that, unlike many existing
utility company, it leads to cost-efficiency in electricity production vehicular communication schemes that utilise V2I communication,
and distribution [2, 3]. By using the advanced metering our scheme uses infrastructure-to-vehicle communication. In our
infrastructure and smart meters, two-way data communication can study, the data in vehicular communication flows from
be achieved between utility companies and consumers. Smart infrastructure (bus stop) to vehicle. However, the proposed scheme
meters are electronic devices that collect and record data on would also allow reverse data flow from a bus to a smart meter
electricity consumption, system status and diagnostic issues. By through a bus stop, e.g. when the utility company wants to
providing information about electricity usage 24/7, smart meters transmit an alert message to its customers such as a message
enable utility companies to have optional pricing programs and containing a reconfiguration file. Thus, the utility control centre
achieve real-time system analysis [2, 4]. can communicate with a smart meter within minutes through a bus
The recently emerging vehicular ad hoc networks (VANET) are a passing by the bus stop.
special form of mobile ad hoc networks. In a VANET, there are (iv) For comprehensive performance results, we conducted
basically two types of communication: vehicle-to-vehicle (V2V) simulations with various densities/distributions of houses and
communication, which is the communication between vehicles, different routing protocols.
and vehicle-to-infrastructure (V2I) communication, which is the (v) We also considered the case when smart meters (houses) are
communication between vehicles and road-side devices. Dedicated located farther than 1000 m from the bus stop, and conducted
short range communication and IEEE 802.11p communication simulations assuming data communication in a hop-by-hop
protocols are used in both communication types. In IEEE 802.11p, fashion. In this scenario, we assumed distant smart meters transmit
the communication range has been expanded to 1000 m and the their data through intermediary houses which relay their smart
data rate has been expanded from 6 to 27 Mbps. The common meter data to the bus stop.
applications of V2V and V2I communication are forward
collision/accident warning, blind spot warning, lane change
warning, do-not-pass warning at the intersection, alerts on 2 Related work
weather-related information and emergency vehicle warning [5–7].
In this work, we explore the existing mechanisms for collecting The main problem in the classical data collecting model from meters,
data from smart meters and offer a new mechanism based on in the traditional electric grid, is that it does not provide two way
vehicular networks. Our solution enables collecting smart meter communications. Due to this problem, utilities cannot learn actual
data using public transportation buses by extending the demand on peak and non-peak hours. In the classical model,
communication capability of smart meters to IEEE 802.11p. Since utilities get their employees to read meters by sending them to

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
& The Institution of Engineering and Technology 2016 515
houses. In this case, there are limited hours to read meters, and if flow of smart meter data from smart meters to buses in this study,
there is no one in a building, since employees cannot enter the the proposed scheme would allow reverse data flow as well. For
building, they cannot read meters. instance, if the utility control centre wants to transmit an alert
In the literature, there are numerous studies on automatic meter message to its customers, e.g. a new electricity unit price or a new
reading (AMR) systems. With recent advances in wireless campaign, or if it wants to send a message containing a
technology, there are studies that expand AMR systems to wireless reconfiguration file, a bus can download this message to the
AMR (WAMR). With their two way communication capability, nearby bus stop within minutes, and finally the bus stop can
AMR systems reduce meter reading expenses of utilities. More transmit the message to the corresponding house. In the same way,
importantly, they help utilities get timely individual electricity regular updates to smart meters could be realised during specific
consumption data from their customers, which helps prevent times of the day.
illegal electricity usage. AMR systems also make it possible to Istanbul has a wide and dense network of bus stops and the
access and control meters even when there is no one in the house. distance between two bus stops in a typical urban neighbourhood
There are four types of AMR systems, given as follows [8]: is around 500 m, as shown in Fig. 2. Note in Fig. 2b the three
neighbouring bus stops in the central Besiktas neighbourhood,
† Drive by AMR systems; located only 465.1 and 402.8 m apart. Hence, in urban
† Fixed network AMR systems; neighbourhoods of Istanbul, smart meters having a WAMR system
† Touch-based AMR systems; with 802.11p capability can easily communicate with a nearby bus
† Walk by AMR systems. stop. With our proposed scheme, houses outside the direct
communication range of a bus stop, i.e. farther than 1000 m from
In [4], the authors offered using the Global System for Mobile a bus stop, e.g. in rural areas far from city centre, can still
Communications (GSM) as the communication medium for communicate with a bus stop through an intermediary house, or
WAMR systems. In the suggested mechanism, meters transmit through multiple intermediary houses, in a hop-by-hop fashion, as
their information, including electricity usage, power quality and shown in Fig. 1b.
outage alarm, to the utility using GSM and at the end of each
month, the utility suggested method would facilitate the generation
of bills and send them to customers via SMS or e-mail. 4 Performance evaluations
In [8], the authors proposed a cost effective novel approach for
AMR systems in rural areas. As an alternative to hiring people for In this section, the performance of the proposed data collecting
meter reading, their approach offers using unmanned vehicles with mechanism is investigated. We conducted our performance
global positioning system capability to read meters. They used the evaluations using the ns-2 [12] simulator with different numbers of
IEEE 802.15.4 communication protocol for their AMR system. sensor nodes. We used the experimentally determined log-normal
They also made simulations for the lifetime of battery used in channel parameters for different smart grid environments based
WSN nodes. According to their performance evaluations, their on [13].
proposed method is well suited for wide areas with few customers. Wireless channel suffers from multi-channel effects, fluctuations
In [9], the authors designed and implemented a WAMR system by in received signal strengths (fading), environmental characteristics
adding wireless modules to existing meters. They used the ZigBee such as outdoor, indoor, and so on and environmental effects such
[10] communication protocol with a short range and low power as noise [14, 15]. Also, there may be diffraction, scattering or
consumption. In their system, wireless modules are connected to reflection on the propagated signal wave. All these factors lead to
meters through RS485 buses. These wireless data collecting modules decreased signal strength at the receiver as the distance between
transmit their data to a sink node either individually or through the source and destination increases [16, 17]. The following
multi-hop communication. Moreover, on the sink node, the received features of radio channels cause the fading phenomenon [18–20]:
data is wirelessly transmitted to the server node via an RS232 bus.
In [11], the authors gave a comprehensive review of the existing † Asymmetrical links: Connectivity between nodes may be
AMR technologies and discussed their challenges. Furthermore, different;
they gave information on future communication technologies for † Non-isotropic connectivity: Connectivity is not required to be the
smart grids, e.g. 3G, DLSM/COSEM and Internet Protocol-based same in all directions;
session initiation protocol (SIP). † Non-monotic distance decay: Nodes far away from the source
Although all the mentioned studies made important amendments may get better connectivity compared with nodes that are
on existing AMR systems or proposed alternative mechanisms for geographically closer.
collecting data from meters, there is no published study about
collecting data from meters using public buses or other public Therefore, a more general wireless channel model, based on
transportation vehicles. Almost all bus stops in Istanbul are smart empirical measurements, is required for more realistic performance
bus stops with embedded computers containing Wi-Fi and other evaluations.
wireless communication modules. In long avenues of Istanbul, or in We have used the log-normal shadowing path loss model as the
any similar city with a large network of smart bus stops, instead of wireless channel model in our simulations, because this model is
hiring people to collect data from meters, meters with wireless known to give more accurate results compared with Nakagami and
capability may transmit their data to the nearest smart bus stop, and Rayleigh models for wireless environments [14]. In the log normal
the buses passing by the bus stop may get the data and transmit it shadowing path loss model, the signal to noise ratio g (d) at a
to a central database owned by the utility company. The proposed distance d from the transmitter is given in as
scheme would be cost effective and provide a novel solution for
transmitting consumer data to the utility company in a timely manner.  

 d
g(d)db = Pt − PL d0 − 10h log10 − Xs − Ph . (1)
d0

3 Proposed scheme In (1), Pt, PL(d0), η, Xσ and Pη denote the transmit power in dBm, the
path loss at a reference distance d0, the path loss exponent, a
In our proposed scheme shown in Fig. 1, data flows from houses, zero-mean Gaussian random variable with standard deviation σ
which have WAMR with 802.11p capability, to a bus stop, and and the noise power in dBm, respectively. The log-normal channel
then collected data on the bus stop is transmitted to a bus which parameters have been obtained with a set of field tests at 5.9 GHz,
drives by on its scheduled time. Upon receiving the collected data, on highway, rural and urban environments. In our simulations, we
the bus transmits it to the utility company using its on-board used 1.61 as the path loss parameter and 3.4 as the shadowing
wide-band communication capability. Although we focus on the deviation.

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
516 & The Institution of Engineering and Technology 2016
Fig. 1 Proposed and extended
a Data collecting scheme
b Scheme allowing hop-by-hop communication

In our simulations, we have modelled a cluster of houses in a [x − r to x + r, y + 2.v − r to y + 2.v + r], and so on. Likewise, the
neighbourhood. For each simulation, we specified the number of starting points for consecutive nodes on the same row are at a
rows and columns, which represent the number of horizontal and distance of h from each other. For instance, if the first node on a
vertical lines along which houses are randomly aligned. Here, the row is placed randomly within the square area [x − r to x + r, y − r
number of rows multiplied with the number of columns in the to y + r], the second one is placed randomly within the square area
cluster gives the number of nodes/houses (see the exemplary [x + h − r to x + h + r, y − r to y + r], the third one placed randomly
clusters in Fig. 3). Although we consider a grid structure, i.e. rows within the square area [x + 2.h − r to x + 2.h + r, y − r to y + r], and
and columns of houses, the houses in the cluster are placed so on. For each grid size, we used 100 different random seed
randomly to emulate houses in a neighbourhood. In our values for the random placement of the nodes on the grid and ran
simulations, the starting point of the first node is fixed and the rest our simulations with these seed values. An exemplary 8 × 8
of the houses are randomly placed on the grid. We developed a cluster, a bus passing by and the bus stop are shown in Fig. 3a. In
C++ program to create the nodes (houses) in the ns-2 format and our performance results, we give the average of all the 100
place them randomly on a grid. Our C++ program takes as input measured values. During each simulation, which lasts for 3900s,
some simulation parameters such as the numbers of rows and seven buses approach the bus stop according to the schedule given
columns in the grid and the initial starting position [x, y] of the in Table 1. In our performance results, we give the average of all
node in the first row of the first column. Further simulation the seven measured values. In order to make our simulations more
parameters are v and h, which denote the vertical and horizontal realistic, we used the actual bus arrival times obtained from the
distances, respectively, between the starting points of consecutive Public Bus Transportation Authority of Istanbul (IETT) for the
nodes, and r which specifies up to how far away a node can be Besiktas Bahcesehir University bus stop, as listed in Table 1.
randomly placed from its starting position both in the horizontal In our simulations, the electricity consumption data is transmitted
and vertical axes. Then, nodes are randomly placed according to from the houses to the bus stop every 10 min, and when a bus arrives
the following rule. Using the initial starting position [x, y], our at the bus stop, it receives from the bus stop the collected electricity
program places the first node (the node at the first row and first consumption data of all the houses in the cluster for the last 10 min
column) randomly within the square area [x − r to x + r, y − r to time period. We utilised constant bit rate (CBR) traffic in our
y + r]. The starting points for the first node in the second row is simulations. As the routing protocol, we tried the Dynamic Source
placed randomly around [x, y + v], within the square area [x − r to Routing (DSR) and ad-hoc on-demand distance vector (AODV)
x + r, y + v − r to y + v + r]. Similarly, the first node in the third routing protocols, since they are the most commonly used routing
row is placed randomly around [x, y + 2.v], within the square area protocols for VANET studies. In our simulations, the buses move

Fig. 2 Wide and dense network of Istanbul bus stop

a Map of Istanbul bus stops
b Distance between neighbouring bus stops

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
& The Institution of Engineering and Technology 2016 517
Fig. 3 Performance evaluations
a 8 × 8 Cluster of nodes
b Node placement scenario
c Node placement scenario for hop-by-hop communication in a low population density neighbourhood

Table 2 Simulation parameters

Parameters name Value

network simulator ns-2

channel model log-normal shadowing
Table 1 Bus arrival times for the Besiktas Bahcesehir University bus number of columns 2–20
stop [21] number of rows 2–20
number of houses 4–400
Bus number Arriving time number of bus stops 1
number of busses 7
29D 10:04:00 AM average maximum bus speed 45–55
43R 10:16:00 AM packet size 100 bytes
25T 10:24:00 AM simulation time 3900 s
58N 10:36:00 AM traffic type CBR
63 10:44:00 AM queue type drop tail
29C 10:55:00 AM routing protocol DSR, AODV
27E 11:05:00 AM vehicle movements same direction with different speed

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
518 & The Institution of Engineering and Technology 2016
Fig. 4 Simulation results in a normal population density neighbourhood for
a Delivery ratio with AODV
b Delivery ratio with DSR
c End-to-end delay with AODV
d End-to-end delay with DSR

in the same direction at a random speed of 45–55 km/h. All the parameters v, h and r. We conducted simulations for
parameters used in our performance evaluations are listed in Table 2. neighbourhoods with up to 400 blocks of houses, with the
We investigated the performance of the proposed scheme for the population densities of 676, 1326 and 169 houses/km2, for normal,
transmission of data packets from smart meters at houses, through high and low population density neighbourhoods, respectively.
the bus stop, to buses, all with IEEE 802.11p capability, using the The proposed data collecting mechanism is ideally offered for
following performance metrics: urban neighbourhoods where consecutive bus stops are <2 km
apart and houses are <1 km away from the nearest bus stop. This
† End-to-end delay: Required time to receive all data packets at will potentially yield a 100% coverage using the IEEE802.11p
destination. protocol. We used a simulation space of size 800 × 800 m in the
† Delivery ratio: Ratio of the number of successfully transmitted maximum to ensure all nodes in the simulation are within 1 km
packets to the total number of transmitted packets. from the bus stop. As noted in Fig. 2, the average distance
between consecutive bus stops in Istanbul is around 500 m in
urban neighbourhoods, and hence an 800 × 800 m simulation
Successful transmission of consumer electricity consumption data space is reasonable. For low, normal and high population density
from all houses to the bus stop is crucial in our mechanism. If data neighbourhoods, we set the distances between consecutive nodes
from some houses is lost during transmission, using the collected to be around 40–80 m by setting the simulation parameters v, h
data the utility company may not produce a realistic data and r accordingly. Hence, along a distance of 800 m, there can be
consumption report, which would decrease the efficiency of load up to 800 m/40 m = 20 houses. We considered row and column
balancing operations. Therefore, achieving a high delivery ratio is sizes of up to 20 in our simulations, resulting in a total number of
important. Similarly, the utility company needs to obtain the up to 400 houses.
electricity consumption data in a timely manner to react to For the scenario when some houses are farther than 1000 m from
changes in load more quickly, which necessitates the data flow the nearest bus stop, we also investigated through simulations the
from the houses, through the bus stop, to the buses to be as fast as performance of a hop-by-hop data communication scheme. In this
possible. Therefore, achieving a minimal end-to-end delay is desired. case, the smart meter data for the distant house is sent to the bus
For the proposed system, we compared the performance of the stop through an intermediary house in a hop-by-hop fashion.
AODV and DSR routing protocols in terms of the delivery ratio
and end-to-end delay performance metrics. Both AODV and DSR
start discovering routes when a demand is initiated by the source 4.1 Performance results for neighbourhoods with
node. Both of them broadcast route request packets to find a path normal population density
from the source to the destination. The main difference between
AODV and DSR is that in DSR multiple route cache entries are For normal population density neighbourhoods, we used the
maintained for each destination, whereas in AODV one route entry simulation parameters v = 40 m, h = 40 m and r = 10 m in our node
is maintained per destination. Furthermore in DSR, source routing placement scenario, which results in a population density of 676
is used whereas in AODV a table-driven routing framework and houses/km2. In this case, the starting points for consecutive nodes
sequence numbers are used to prevent loops [22–24]. are placed 40 m apart on both the horizontal and vertical axes. In
For testing the performance of the proposed scheme for varying Fig. 4, we give the simulation results for delivery ratio and
node population densities, we performed simulations with varying end-to-end delay with the AODV and DSR routing protocols.
numbers of nodes, representing neighbourhoods with different With AODV, for lower numbers of nodes the delivery ratios are
population densities. We achieved this by changing the simulation higher, and all delivery ratios are between 98.00 and 100%. As the

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
& The Institution of Engineering and Technology 2016 519
Fig. 5 Simulation results in a high population density neighbourhood for
a Delivery ratio with AODV
b Delivery ratio with DSR
c End-to-end delay with AODV
d End-to-end delay with DSR

numbers of rows and columns increase, delivery ratio decreases. (for the grid 20 × 18) and 75.74%, respectively. With DSR, for
Especially, when both the number of rows and the number of most cases the delivery ratio is between 96.00 and 100%. It starts
columns are higher than 15, the delivery ratio goes under 40%. decreasing when the number of rows and columns are both higher
The minimum and average delivery ratios with AODV are 9.39% than 18, however it stays between 82.28 and 96.00%. The

Fig. 6 Simulation results in a low population density neighbourhood for

a Delivery ratio with AODV
b Delivery ratio with DSR
c End-to-end delay with AODV
d End-to-end delay with DSR

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
520 & The Institution of Engineering and Technology 2016
Fig. 7 Hop-by-hop communication with the AODV and DSR routing protocols in low population density neighbourhoods for
a Delivery Ratio results
b Average end-to-end delay results

minimum delivery ratio of 82.28% occurs for the 17 × 19 grid and neighbourhood, which results in a population density of 1326
the average delivery ratio for all the simulated grid sizes is houses/km2. In this case, the starting points for consecutive nodes
99.61%. We can see that for larger grid sizes, i.e. higher numbers are placed 20 m apart on the horizontal axis for nodes on the same
of nodes, the delivery ratios are dramatically low with AODV, and row and 40 m apart on the vertical axis for nodes on the same
in general, the delivery ratios are much better with DSR. column. Note here that we allowed the spacing between consecutive
As seen in Fig. 4, with both routing protocols, when the number of rows to be larger than the spacing between consecutive columns to
nodes increases, the end-to-end delay also increases. The delay allow for interior parallel roads between rows of houses.
values are higher with AODV, compared with DSR. This may be We give the simulation results for delivery ratio and end-to-end
due to the fact that the delay increases when the routing table delay, with both the AODV and DSR routing protocols, in Fig. 5.
entries in AODV expire. Note that a routing table entry expires if The average delivery ratio for all the number of nodes is 73.47%
it has not been used recently. All in all, DSR gives better results for AODV and 99.88% for DSR. With AODV, when both the
with the maximum delay of only 0.56 ms, compared with AODV number of rows and columns are higher than 16, the delivery ratio
with the significantly larger maximum delay of 135.31 ms. The drops under 35% and the minimum delivery ratio is 9.95% (for
average delay values are 31.68 and 0.08 ms with AODV and DSR, the 20 × 19 grid). With DSR, the minimum delivery ratio of
respectively. Note that in our proposed method most of the nodes 91.99% occurs for the 18 × 20 grid. The maximum observed delay
are static and the only mobile nodes are the buses. Since the values are 111.94 and 9.07 ms, and the average observed delay
number of mobile nodes in our network is small, DSR gives better values are 30.01 and 3.60 ms, with AODV and DSR, respectively.
results. Furthermore, since AODV uses route expiry, packet drops In terms of delivery ratio and delay, while AODV is suitable for
may occur in this protocol when a new route is discovered, only smaller grids with lower number of nodes, DSR can be used
resulting in increased packet transmission times. for all grid sizes.

4.2 Performance results for neighbourhoods with high 4.3 Performance results for neighbourhoods with low
population density population density

We used the simulation parameters v = 40 m, h = 20 m and r = 10 m For low population density neighbourhoods, we used the simulation
in our node placement scenario for a high population density parameters v = 80 m, h = 80 m and r = 40 m in our node placement

Fig. 8 Simulation results for

a Delivery ratio
b End-to-end delay

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
& The Institution of Engineering and Technology 2016 521
scenario, which results in a population density of 169 houses/km2. In applications, health applications, environmental applications, home
this case, the starting points for consecutive nodes are placed 80 m applications, smart grid environments and vehicular
apart both on the horizontal and vertical axes for nodes placed on communication. In this work, we merge WSNs with VANETs in a
the same row and column, respectively. novel scheme for solving the data communication problem in
The simulation results for delivery ratio, with both the AODV and smart grids. We have offered a new data collecting mechanism for
DSR routing protocols, are given in Fig. 6. As seen in the figure, with smart meters by extending their communication capability to IEEE
AODV, for lower number of nodes, the delivery ratio is 100%, and 802.11p. Changing their communication model, we propose
for the number of nodes between 50 and 80, the delivery ratio is collecting smart grid data from smart meters by using public
between 98.00 and 100%. As the number of nodes increases above transportation buses with wireless communication capability. In
80, the delivery ratio sharply decreases. Particularly, for the our proposed scheme, smart meters transmit their data to a bus
number of nodes higher than 90, the delivery ratio decreases down stop, and then the bus stop transmits the data to a passing-by bus,
to 90%. Although the minimum delivery ratio is 87.95% (for the which in turn transmits it to a central server of the utility company.
grid 10 × 10), the average delivery ratio for all grid sizes is We evaluated the performance of our proposed scheme and made
99.34%. With DSR, for all number of rows and columns, the simulations with two different routing protocols, namely AODV and
delivery ratio is 100%. In terms of delivery ratio, the proposed DSR, for obtaining end-to-end delay and delivery ratio values. The
mechanism gives good results with both AODV and DSR for channel parameters used in our simulations were obtained from a
small grid sizes, however DSR is more desirable for larger grid sizes. set of field tests at 5.9 GHz in different environments. Our
In Figs. 6c and d, the simulation results with both AODV and DSR performance evaluations show that the proposed data collecting
are given for end-to-end delay. With both protocols, when the mechanism achieves significantly better delivery ratio and lower
number of nodes increases, the delay also increases. For number delay when DSR is used.
of nodes higher than 80, the delay exceeds 20 ms and the We identify possible future research directions as follows:
maximum delay is 43.01 ms (for the 10 × 10 grid and with
AODV). The maximum delay value with DSR is only 0.06 ms. † We transmitted smart meter data of size 100 bytes every 10 min
The average delay values are 3.28 and 0.037 ms with AODV and from houses to a bus stop. The influence of varying packet sizes
DSR, respectively. and transmission periods on the efficiency of the proposed scheme
may be studied.
4.4 Performance results for hop-by-hop communication † We did not consider the case when a bus would not arrive at the
in low population density in rural neighbourhoods bus stop on its scheduled arrival time. The no show case for a bus
may be studied and possible countermeasures could be investigated.
We investigated the performance of the proposed scheme also with † For houses located more than 1000 m away from the nearest bus
hop-by-hop communication for low population density stop, we proposed hop-by-hop communication and made
neighbourhoods. We did simulations with both the AODV and preliminary simulations. A more comprehensive study could be
DSR routing protocols. We used two 10 × 10 grids representing conducted on this scenario for a detailed analysis of hop-by-hop
two consecutive neighbourhoods, as shown in Fig. 3c, with the communication.
simulation parameters v = 80 m, h = 80 m and r = 40 m. In this † The security and privacy aspects of the proposed scheme may be
scenario, we tried to achieve coverage for houses which are more studied. Efficient mechanisms may be investigated for
than 1000 m farther from the bus stop. These houses transmit their communicating smart grid data in a secure and privacy preserving
data to the bus stop through another house located at the manner in the proposed scheme.
neighbouring grid that is within the communication range of the † In the proposed scheme, the IEEE 802.11p communication
bus stop. We conducted simulations for only two neighbouring protocol, which supports a communication range of up to 1000 m,
grids, each covering an area of ∼1 km2, however this scenario is used. Different wireless technologies may be investigated for the
could be extended to cover multiple neighbourhoods and longer proposed system and their influence on the effective
distances where houses would transmit their data to the bus stop communication distance may be studied.
through multiple intermediary neighbourhoods in a hop-by-hop † In the proposed scheme, one smart meter per building is assumed.
fashion. We repeated all the simulations for 100 times and took For buildings with multiple smart meters, one smart meter
the averages for the delivery ratio and end-to-end delay values. (preferably on the top floor of the building) could be picked as the
Fig. 7 presents the simulation results for delivery ratio and central smart meter which is responsible for collecting data from
end-to-end delay with both AODV and DSR. Obviously, both the other smart meters in the building using a wired or a wireless
delivery ratio and delay values are much better with DSR. While a communication technology, then using 802.11p the accumulated
87.72% delivery ratio is achieved in 56.84 ms with AODV, a data could be sent by this central smart meter to buses. The
100% delivery ratio is achieved in only 0.56 ms with DSR. alternative would be individual data transmissions from all smart
meters in a building to buses. Such scenarios extending the
capability of the proposed scheme could be explored, and their
4.5 Overall performance analysis efficacy and performance could be investigated.
† If buses are passing by an area with high-rise buildings, data
In Fig. 8, we summarise and compare our simulation results for the transmission from buildings in the back streets may be hindered,
proposed smart grid data collecting scheme with both the AODV and which could be mitigated by using multi-hop communication. The
DSR routing protocols, and for low, normal and high population influence of unusual or problematic scenarios on the effectiveness
density neighbourhoods. As seen in the figure, DSR results in a of the proposed scheme could be investigated and possible
higher average delivery ratio than AODV for all population countermeasures could be studied.
densities. Note that the average delivery ratio with DSR is equal to † Potential other uses of the proposed technology could be
or only slightly <100% for all population densities. Similarly, explored, e.g. remote patient monitoring for healthcare
DSR results in significantly lower average end-to-end delay applications, where patient data, measured by sensors, is
compared with AODV. Note that in high population density communicated to a physician periodically.
neighbourhoods, the average end-to-end delay with DSR is orders
of magnitude lower than that with AODV.
6 References

5 Conclusions 1 Akyildiz, I.F., Su, W., Sankarasubramaniam, Y., et al.: ‘Wireless sensor networks:
a survey’, Comput. Netw., 2002, 4, (38), pp. 393–422
2 Gungor, V.C., Lu, B., Hancke, G.P.: ‘Opportunities and challenges of wireless
With recent advances in WSNs, WSNs have been deployed widely sensor networks in smart grid’, IEEE Trans. Ind. Electron., 2010, 57, (10),
in many applications, e.g. military applications, monitoring pp. 3557–3564

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
522 & The Institution of Engineering and Technology 2016
 3 Erol-Kantarci, M., Mouftah, H.T.: ‘Wireless sensor networks for domestic energy 14 Rappapport, T.: ‘Wireless communications: principles and practice’ (Prentice Hall,
management in smart grids’. Proc. Communications (QBSC), Kingston, Canada, 2002, 2nd edn.)
May 2010, pp. 63–66 15 Nabar, R.U., Bolcskei, H., Kneubuhler, F.W.: ‘Fading relay channels: performance
4 Sheikh, S.S., Sharma, S.: ‘Design and implementation of wireless automatic meter limits and space–time signal design’, IEEE J. Sel. Areas Commun., 2004, 22, (6),
reading system’, Int. J. Eng. Sci. Technol., 2011, 3, (3), pp. 2329–2334 pp. 1099–1109
5 Bilgin, B.E., Gungor, V.C.: ‘Performance comparison of IEEE 802.11p and IEEE 16 Zuniga, M., Krishnamachari, B.: ‘An analysis of unreliability and asymmetry in
802.11b for vehicle-to-vehicle communications in highway, rural, and urban areas’, low-power wireless links’, ACM Trans. Sensor Netw., 2007, 3, (2), pp. 1–30
Int. J. Veh. Technol., 2013, 2013, pp. 1–10 17 Bilgin, B., Gungor, V.C.: ‘Adaptive error control in wireless sensor networks under
6 Molisch, A.F., Tufvesson, F., Karedal, J., et al.: ‘A survey on vehicle-to-vehicle harsh smart grid environments’, Sens. Rev., 2012, 32, (3), pp. 203–211
propagation channels’, IEEE Wirel. Commun., 2009, 16, (6), pp. 12–22 18 University of California: ‘SCALE: a tool for simple connectivity assessment in
7 Gerla, M., Kleinrock, L.: ‘Vehicular networks and the future of the mobile lossy environments’ (Center for Embedded Networked Sensing, University of
internet’, Comput. Netw., 2011, 55, (2), pp. 457–469 California Los Angeles, 2003), pp. 1–16
19 Govindan, R., Zhao, J.: ‘Understanding packet delivery performance in dense
8 Tuna, G., Gungor, V.C., Gulez, K.: ‘Unmanned vehicle-aided automated meter
wireless sensor networks’. Proc. First Int. Conf. on Embedded networked sensor
reading’. Proc. Int. Conf. Broadband Communications and Biomedical
systems, Los Angeles, USA, November 2003, pp. 1–13
Applications, Melbourne, Australia, November 2011, pp. 289–293
20 University of California: ‘Complex behavior at scale: an experimental study of
9 Miao, D., Xin, K., Wu, Y., et al.: ‘Design and implementation of a wireless low-power wireless sensor networks’ (Computer Science Department, University
automatic meter reading system’. Proc. Int. Conf. on Wireless Communications of California Los Angeles, 2002), pp. 1–11
and Mobile Computing: Connecting the World Wirelessly, Leipzig, Germany, 21 ‘Bus Line Detail’. Available at http://mobil.iett.gov.tr/sa/mobil/hatarama/hatdetayi,
June 2009, pp. 1345–1349 accessed 12 December 2014
10 ‘ZigBee Alliance’. Available at http://www.zigbee.org, accessed 01 December 2014 22 Kaushik, S.S., Deshmukh, P.R.: ‘Comparison of effectiveness of AODV, DSDV,
11 Khalifa, T., Naik, K., Nayak, A.: ‘A survey of communication protocols for automatic and DSR routing protocols in MANETs’, Int. J. Inf. Technol. Knowl. Manag.,
meter reading applications’, IEEE Commun. Surv. Tutor., 2011, 13, (2), pp. 168–182 2009, 2, (2), pp. 499–502
12 ‘The Network Simulator - ns-2’. Available at http://www.isi.edu/nsnam/ns/index. 23 Yadav, N.S., Yadav, R.P.: ‘The effects of speed on the performance of routing
html, accessed 01 October 2014 protocols in mobile Ad-hoc networks’, Int. J. Electron. Circuits Syst., 2009, 1,
13 Kunisch, J., Pamp, J.: ‘Wideband car-to-car radio channel measurements and (2), pp. 79–84
model at 5.9 GHz’. Proc. Vehicular Technology Conf., Calgary, Canada, 24 Rasheed, M.R., Khan, M.K., Naseem, M., et al.: ‘Performance of routing protocols
September 2008, pp. 1–5 in WiMAX networks’, Int. J. Eng. Technol., 2010, 2, (5), pp. 412–417

IET Intell. Transp. Syst., 2016, Vol. 10, Iss. 8, pp. 515–523
& The Institution of Engineering and Technology 2016 523